Full opinion text
MEMORANDUM OPINION AND ORDER SHERMAN G. FINESILVER, District Judge. Plaintiffs and the class they represent allege that defendant Martin Marietta Corp. discriminated' against them because of their race, sex, color, or national origin thereby denying them equal employment opportunities in violation of federal law. The claims relate to defendant’s promotional policies. There are also allegations by two of the named plaintiffs (Ms. Rich and Mr. Tafoya) that defendant has retaliated against them for their having filed charges with the Equal Employment Opportunity Commission. The claims, and this court’s jurisdiction, arise under Title VII of the Civil Rights Act of 1964, as amended, 42 U.S.C. §§ 2000e through 2000e-17 (1976) as well as under 42 U.S.C. § 1981 and 28 U.S.C. § 1343(4) (1976). The seven named plaintiffs are black female (Ms. Rich), black male (Mssrs. Franklin, Langley, Collier, Craig, and Chappel), and Hispano-American male (Mr. Tafoya) citizens of the United States, all of whom reside in the district of Colorado. The class they represent includes all female, black, and Hispano-American employees of Martin Marietta’s Waterton (Colorado) facility as of May 11, 1971. By stipulation of the parties, the trial was bifurcated. The instant trial was had on the issue of liability alone and was further limited to the issue of liability solely as to the seven named plaintiffs. Broad based discovery has, nonetheless, gone forward and plaintiffs presented plant-wide statistical evidence in support of their individual claims. We have carefully considered all the evidence adduced at both trials (the matter has been before the court previously) and have reviewed the exhibits and all other matters of record. With the exception of plaintiffs Rich and Chappel, none of the plaintiffs has met their initial burden of establishing a prima facie case of employment discrimination under Title VII. Defendant has rebutted Ms. Rich’s prima facie case but has not rebutted Mr. Chappel’s. We conclude, therefore, that there is Title VII liability towards Mr. Chappel alone of all the named plaintiffs. None of the plaintiffs has succeeded in establishing a case of employment discrimination under 42 U.S.C. § 1981. We thus find in defendant’s favor as to six of the plaintiffs, find liability against defendant as to one of the plaintiffs, and will determine Mr. Chappel’s remedy in subsequent proceedings. Our findings and conclusions follow: I This lawsuit had previously been before the court in a trial which resulted in a judgment for defendant on all counts. That judgment in the earlier trial was reversed and the cause remanded to this court, with instructions, for further proceedings. Rich v. Martin Marietta Corp., 522 F.2d 333 (10th Cir. 1975). The Circuit Court reversed and remanded for reconsideration of plaintiffs’ request for additional discovery, for reconsideration of the question of class certification, and for reevaluation of plaintiffs’ claims. All of plaintiffs’ discovery requests have now been granted. Defendant has spent a large amount of time and money developing and collecting statistical information for plaintiffs. All statistics have been adjusted so as to exclude American Indians and Oriental males. See, Rich, supra, 522 F.2d at 337, 343, 346, and 346 n.12. The class has been certified according to plaintiffs’ request. Having already conducted full discovery, the parties have agreed to holding separate trials on the individual and class claims. Trial on the issue of liability as to the class will, accordingly, be set in due course. The court has reevaluated the plaintiffs’ claims. In so doing, we have relied upon the evidence which was presented at the first trial and that which has now been adduced at the second trial. See, Rich, supra, 522 F.2d at 347, 349. We have considered the claims anew and have made findings and conclusions in light of the comments and guidance contained in the Tenth Circuit’s opinion. Before weighing the supplemental evidence adduced at the second trial, it would be well briefly to review the state of the evidence as it stood at the end of the first trial. This is a promotions case. At issue are various of defendant’s promotional policies. Defendant has more than one promotion policy because it has more than one class of employee. Defendant, an aerospace manufacturer employing upwards of 5,000 persons at its Waterton plant, groups its employees into three relevant classes. Those categories are “salaried,” “hourly-out-of-unit,” and “hourly-in-unit.” The salaried employees are professionals, mostly engineers, together with officials or managers. This group accounts for roughly half of defendant’s labor force. The salary grades are designated numerically from 40 to 50. The lowest salary grade is 40, the highest is 50. These employees are promoted from one grade to another strictly on the basis of merit as determined by a “totem pole” evaluation system. Under this system, salaried employees are competitively rated and are ranked, in order of their performance, relative to one another. A salaried employee’s position on the totem pole is highly determinative of the employee’s likelihood of promotion. The hourly out-of-unit employees include technicians, office and clerical workers, craft (skilled) and operative (unskilled) workers. The hourly out-of-unit grades are designated numerically from 12 to 1, and alphabetically, A and B. The lowest hourly out-of-unit grade is 12, the highest is A, with the normal progression being through the numerical grades to the letter grades. Promotions of these hourly technicians through their own ranks is based upon a combination of qualifications and seniority. A promotion in this category is first offered to the most senior qualified employee in the job family group. A second sort of promotion, a promotion out of the hourly ranks and into the professional category (that is, from “hourly to salary”) is also a possibility. Hourly to salary promotions are made strictly on the basis of merit. Hourly in-unit employees constitute the third category. These employees are represented by a collective bargaining unit. Promotions in this category are awarded solely on the basis of seniority. The seven plaintiffs are all long term employees of defendant corporation. As noted, they are black female (Ms. Rich), black male (Mssrs. Franklin, Langley, Collier, Craig, and Chappel), and HispanoAmericans (Mr. Tafoya). They represent varied job progression ladders and careers. Ms. Rich is an engineer, Mr. Franklin an accountant, Mr. Langley a developer, Mr. Collier a millwright, Mr. Craig a millwright, Mr. Chappel an electrician, and Mr. Tafoya an electronic developer. Over the broad period of time involved here, their specific job titles varied as their responsibilities rose or fell as a consequence of promotions, demotions, or reclassifications. As a group, and during most of the relevant time period, the seven plaintiffs were generally either salaried or hourly out-of-unit employees. Those who were professionals (salaried) claim that they were not promoted as rapidly as other, white male, salaried co-workers. Those who were technicians (hourly out-of-unit) contend that they were neither promoted within the technician ranks as rapidly as other, white male, hourly co-workers nor offered the desirable promotions into the professional ranks (hourly to salary promotions) as were the other workers. All allege that this asserted difference in treatment is the result of defendant’s use of the impermissible factors, race, sex, color, or national origin. They allege that the evaluation systems used by defendant to determine promotions were impermissibly subjective and that such subjectivity tainted the ratings. Each plaintiff filed a charge with the Equal Employment Opportunity Commission in the latter part of 1969. The instant lawsuit, initially filed on May 11, 1971, resulted. In our earlier, unpublished ruling we made extensive findings of fact of which the foregoing recitation is but a part. We will not retravel ground already covered. The Tenth Circuit, in an appendix to its opinion, summarized our findings as to each of the individual plaintiffs. For the present purposes, we too shall use the Tenth Circuit’s summary as our own. We incorporate it as Appendix I to this opinion. We earlier held, for reasons noted, that the evidence, as it then stood, was insufficient to make out a prima facie case. We were not convinced, on the record then before us, that plaintiffs’ statistical evidence showed any discrimination in promotions nor were we convinced that the plaintiffs had shown themselves to be qualified for promotions to available positions. Plaintiffs were, however, afforded an overly limited range of discovery in the first trial. Such limitations tended to frustrate the plaintiffs’ efforts at the earlier trial. The instant, second trial has proceeded after wide ranging discovery. Both sides have introduced supplemental evidence. Some of this new evidence relates directly to the non-statistieal side of the lawsuit and so directly compliments the evidence of the first trial. Most, indeed the great majority, of the new evidence is statistical, however, and (because of the greater scope of discovery) stands wholly independently of the statistical evidence adduced at the first trial. Because of the defects of the statistics of the first trial, and because of the complete and sophisticated statistical presentation made at the second trial, we disregard entirely all earlier statistics. With this basic outline of events as background, we now proceed to analyze plaintiffs’ various allegations in some detail, weighing the crucial evidence on each issue. We include in our analysis appropriate evidence from each of the two trials. II Plaintiffs allege that defendant denied them promotions, impermissibly discriminating on the basis of race, sex, color, or national origin in violation of Title VII and of 42 U.S.C. § 1981. Plaintiffs’ general theory of the case is that they each were qualified for promotion, that there were available positions into which they could have been promoted, and that there is evidence of discrimination in defendant’s promotion policies from which it can be inferred that these qualified plaintiffs were denied available promotions because of their sex, race, national origin, or color. The massive volume of evidence may be logically clustered into three general groups, paralleling plaintiffs’ claims, as follows: (a) evidence of qualifications for promotion, (b) evidence of available positions, and (c) evidence (statistical and otherwise) tending to show the presence or absence of an inference that defendant’s promotional policies had a discriminatory intent or impact. We treat separately each area and, in turn, summarize the more important evidence. Where conflict exists, we note our findings of fact on such issues. A Plaintiffs allege that they were qualified for promotion. Defendant disputes this allegation. Much, indeed most, of the first trial consisted of testimony pertaining to the individual qualifications and work experience of the several plaintiffs. The Tenth Circuit’s summary of our earlier findings is attached hereto as Appendix I. In brief, it may be said that plaintiffs have each made a showing of lengthy tenure and of certain levels of experience. Plaintiffs have offered no supplemental evidence concerning qualifications, resting instead on the record developed at the first trial. Defendant, for its part, has offered supplemental evidence, both general and specific, which expands upon its assertion that these plaintiffs were simply not qualified to receive promotions sooner than they, in fact, received them. Generally, defendant offered additional evidence to show that seniority was not a particularly relevant factor in awarding promotions such as these plaintiffs sought. As salaried and hourly out-of-unit employees, these plaintiffs were in line for promotions of three kinds: salaried, hourly out-of-unit, and hourly to salary. The evidence remains undisputed that salaried and hourly to salary promotions are determined solely on the basis of merit. While seniority does play a part in determining hourly out-of-unit promotions, performance plays at least as great a part. Specifically, defendant’s supplemental evidence further explained the hourly out-of-unit structure (including definition of the A and B grade levels), the hourly to salary promotion trends (including an analysis of labor pools), and, finally, offered some further explanation as to two of the plaintiffs (Ms. Rich and Mr. Collier). With respect to the hourly out-of-unit structure, defendant’s supplemental evidence expanded upon evidence already in the record. As noted, the grades ascend from 12 to 1 and then from B to A. Hourly out-of-unit employees include technicians as well as office and clerical, skilled, and unskilled workers. The technicians are, predominantly, in the higher grades. The highest grades are, of course, the A and B levels. There was testimony as to the distinction between A and B level technician job tasks. Mr. Charles A. Hall (a supervisor in the engineering propulsion laboratory) stated that engineering technician A’s were qualified to supervise other technicians, to do engineering work such as fluid flow or instrument calibration, or to do planning. An A technician was expected to be able to set up, run, and command a test in a safe manner. Mr. Thomas Johancen (a supervisor and a Millwright A) testified that the A millwright was qualified to overhaul pumps, compressors, cranes, boilers, and exhaust systems. Mr. Johancen further testified that the A electrician was qualified to work with wiring schematics, conduits, and electrical codes in the installation of compressors, pumps, and new machines, and in repair of electric motors. In all cases, the B level technician generally assisted the A level or performed work calling for less initiative, skill, and experience. With respect to the hourly to salary promotion trends, defendant’s supplemental evidence tended to show that the higher hourly out-of-unit grades were both more qualified for, and more likely to get such promotions. Defendant’s exhibit BBBB notes that, using data as of December 31, 1967 as an example, most (that is, 95%) of the A and B level workers were performing jobs with high technical requirements. By comparison, only 12% of the employees in hourly labor grades 1-6 and only 3% of the employees in grades 7-12 were in technical jobs. As previously noted, the salaried positions were professionals, for the most part engineers doing work of a technical nature. Dr. Roy Yamahiro (a manager responsible for performance evaluations) testified that, for an A or B level technician, promotion from hourly to salary was a normal movement upward. This promotion would represent a continuation and advancement of their already developed skills and progressively increasing initiative and responsibility. Dr. Yamahiro noted that, for the next several grades of hourly workers (grades 1-6), promotion from hourly to salary was not a normal movement upward. Those in this group who were promoted to salary generally advance as a supervisor or as an official or manager. Dr. Yamahiro’s testimony was based upon a personal review of every single hourly to salary promotion, by name and date, from 1965 through 1971. Defendant characterized such unrefuted supplemental evidence on this score as establishing relevant “labor pools” with respect to hourly to salary promotions. Defendant referred to the A and B grades as the high pool, grades 1-6 as the medium pool, and grades 7-12 as the low pool. Finally, with respect to two of the plaintiffs, defendant introduced supplemental evidence to explain and clarify matters brought up at the first trial. Mr. Robert Anschicks (a supervisor for whom Ms. Rich had worked) further explained why, in his opinion, the plaintiff Ms. Rich was not qualified for promotion from salary grade 43 to salary grade 45. He noted that, though she had some ten years’ experience, it was not progressive and that she had poor knowledge of load paths. He concluded that she was not technically qualified for promotion and did not have the proper work habits. On recross examination, Mr. Anschicks explained that he had recommended Ms. Rich for advancement in 1968 in the hope that more challenging work might somehow increase her interest and improve her poor attendance record. This last matter is referred to in Rich, supra, 522 F.2d at 349-50 (Appendix A). Mr. Johancen testified that the plaintiff Mr. Collier was not offered a promotion from hourly grade B to A in 1969. Referring to Defendant’s Exhibit U from the first trial, the witness described that January 24, 1969 document as a reclassification review. Mr. Collier, through his response on that form, had stated that he was not then interested in going to Millwright A. The reclassification review form was not an offer of promotion but was, apparently, a form used, generally, to ascertain employee preferences. Defendant did not thereby necessarily indicate that it deemed Mr. Collier to be qualified for promotion at any time prior to June of 1972 (when he was, in fact, promoted). This testimony relates to matters alluded to in Rich, supra, 522 F.2d at 353 n.2 (Appendix A). Based upon the evidence of the first trial and the supplemental evidence adduced at the second trial, we make the following findings of fact with respect to qualifications: Length of experience was not a major factor in the relevant promotional policies of defendant. Although plaintiffs were, in no case, the most qualified persons for promotion, they did show generally that they had had lengthy tenure and had gained certain experience working for the defendant company. We find that there is a significant difference in qualifications between A and B level technicians. We further find that there are relevant labor pools for hourly to salary promotions. The high pool includes hourly out-of-unit grades A and B, the medium pool includes grades 1-6, and the low pool includes grades 7-12. Finally, we find that defendant did not consider Ms. Rich to be technically qualified for promotion in 1968 and that defendant did not consider Mr. Collier to be qualified for promotion in 1969. B Plaintiffs allege that there were available positions into which they could have been promoted. Defendant vigorously contests this allegation. Plaintiffs’ theory of the case is that the asserted violations are _ continuing violations, not requiring any showing of specific promotions within 90 or 210 days of plaintiffs’ having filed charges with the EEOC. Defendant argues that plaintiffs must point to specific promotion openings within 210 days preceding the filing of charges with the EEOC. Plaintiffs have offered no supplemental evidence concerning the availability of positions into which they could have been promoted. They rest upon the record of the first trial. Defendant, on the other hand, did introduce supplemental evidence. Mr. Leland Gillespie (a manager in compensation, wage, and salary) testified to the absence of relevant hourly out-of-unit promotions within the 210 days prior to Mssrs. Tafoya, Langley, Craig, and Collier’s having filed charges with the EEOC. Mr. Anschicks testified that there were no relevant salaried promotions during the 210 days before Ms. Rich filed charges with the EEOC. None of this testimony was countered by plaintiffs nor was it seriously undermined on cross-examination. We find that defendant’s testimony on this score was reasonable, credible, and worthy of belief. We, accordingly, find that there were no relevant, promotions during the 210 days before the dates on which Ms. Rich, Mr. Tafoya, Mr. Langley, Mr. Craig, and Mr. Collier filed charges with the EEOC, C Plaintiffs allege that defendant’s promotional policies have had a discriminatory intent or impact. Defendant denies that this is the case. The evidence on this score is both statistical and non-statistical. Because of the importance of the statistical evidence in this law suit, we shall discuss it separately from the non-statistical evidence. Non-Statistical Evidence Both plaintiffs and defendant have supplemented the circumstantial, non-statistical evidence of the first trial. Plaintiffs’ supplemental evidence relates to the company’s evaluation system (which they contend to be impermissibly subjective), to events surrounding a certain public demonstration, and to various tabulations of minority positions at the defendant company. Defendant has responded. In addition, defendant has adduced supplemental evidence with respect to its affirmative action plan. As to the evaluation system, plaintiffs’ witness, Mr. John Smith (an acting section chief in the library) testified that, though he personally gave fair ratings, he thought that he could have been unfair if he had wanted to be. The. witness stated that, though he had complied with certain of defendant’s written directions for preparation of evaluations, he did not have formal instructions on how to fill out the forms. Mr. Smith testified that he was cognizant of the fact that Martin Marietta had an affirmative action plan during the period 1969-70 (the witness thought that this was the period during which he had acted as a supervisor in the library). He described the plan as a “numbers game” which was being pushed by upper management to increase the numbers of minority employees. Defendant’s witness, Dr. Yamahiro, qualified as an expert in the field of employee performance appraisals, stated that the evaluation forms used by defendant were, in his opinion, more objective than the forms used internally by the Equal Employment Opportunity Commission or by the Department of Justice. He noted that the reason defendant’s forms might have some element of subjectivity involved is due to the nature of Martin Marietta’s business. Where jobs are not easily definable and where job tasks are not repetitive or redundant, it becomes harder to be wholly objective. Dr. Yamahiro testified that defendant had sought to increase objectivity in its ratings by designing the forms so that a rating official was required, first, to describe the actual job assignment of the employee being rated and, then, to measure the employee’s performance against his particular job tasks. The witness stated that defendant had also sought to increase objectivity by ensuring that a second official review the evaluations prepared by the initial rating official. Based upon his review of defendant’s written evaluation policy, the witness concluded that the evaluations were objective. As to the public demonstration, Mr. Tommy Franklin (one of the plaintiffs) testified that he had asked for a vacation day in order to participate in a civil rights demonstration against Martin Marietta which was scheduled to take place during his working hours. Mr. Franklin did not get a vacation day, took the day off anyhow, and received a written reprimand as a result. Mr. John Smith testified that he knew of the planned demonstration in advance and that he had been told to make sure that all of his employees reported for work on the day in question. He stated that, following the demonstration, the defendant company investigated to determine whether employees who had called in sick had, instead, participated in the demonstration. Defendant’s witness, Mr. Bobby Frank Leonard (a manager in employee relations) testified that reprimands had issued to white and black employees alike who, without prior authorization, had not reported to their regularly scheduled work shift because of the demonstration. As to minority positions in the company, plaintiffs offered several tabulations. Exhibits 107 and 109 are the more significant of the exhibits in this category. Exhibit 107 demonstrates the average time spent in grade as an Electrician B prior to promotion to Electrician A. While HispanoAmericans averaged 18 months and whites averaged 24.3 months in grade, the exhibit reflected that blacks averaged 52.6 months in grade. As previously noted, see Appendix I, infra, plaintiff Chappel was an Electrician B for the entire relevant period 1965 through 1971. Exhibit 109 demonstrates the average time spent in grade as a Millwright B prior to promotion to Millwright A. Whites averaged 14.3 months, HispanoAmericans 16.3 months, yet blacks averaged 68.3 months in grade. As previously noted, plaintiffs Craig and Collier were both Millwright B’s at various times between 1969 and 1971. Defendant has tendered modified exhibits corresponding to plaintiffs’ exhibits 107 and 109. The modified exhibits are adjusted so as to exclude post-1971 data. Modified 107 shows average time in grade for Electrician B as 24.3, 12, and 37.3 months for white, Hispano-American, and black employees, respectively. Modified 109 shows average time in grade for Millwright B as 14.2, 17.5, and “_” months for white, Hispano-American, and black workers, respectively (there were no black Millwright B’s who were promoted to A during 1966-69). Plaintiffs’ other tabulations included exhibits showing accounting department employees with college degrees such as plaintiff Franklin had, as well as other exhibits which showed salary levels, workforce composition, and other background data on defendant company. Defendant likewise introduced several tabulations. The more significant of these were defendant’s series FFFF1 to 7 and GGGG1 to 14. Series FFFF compared each individual plaintiff’s promotion history during the relevant period 1966-69 against the promotion history of white males who had been initially hired during or in close proximity to the starting dates of the plaintiffs and into the same or similar initial job position and labor grade. Series GGGG compared the individual plaintiffs against white males who happened to be in comparable situations as were plaintiffs at each relevant promotion point. These exhibits sought to match the plaintiffs with white males who had the same degree of lengthy tenure and experience as did the plaintiffs. Defendant’s witness, Dr. Gollob, testified that these exhibits showed the individual plaintiffs to be generally in the mid range, receiving neither more nor less promotions than the majority of comparable white males. He concluded that these exhibits did not support any claim of discriminatory promotions. Defendant’s other tabulations included exhibits showing workforce composition, salary levels, and other background information on defendant company. The court excluded certain of those exhibits as being irrelevant. Plaintiffs repeatedly represented, and the record is absolutely clear, that this is a promotions case. It is not a hiring case nor is it a disparate pay case. Finally, in the category of non-statistical circumstantial evidence, defendant introduced supplemental evidence concerning its affirmative action plan. Mr. Leonard testified that the defendant company had, in late 1967 or early 1968, determined to do something about the fact that it had no black Millwright A’s. The company tried, without success, to recruit black Millwright A’s in 1968. The company sought to promote its black Millwright B’s and established a special training program in 1968 in order to qualify black Millwrights for promotion. Mr. Leonard testified to defendant’s affirmative action plan in general, noting that his office had reviewed every promotion during the period commencing in 1966 and had checked, in each instance, to see whether there were any qualified minorities who could fill the job. The witness stated, further, that defendant company had never had a policy of segregating its job classifications. By way of explanation, the witness noted that prior to 1965 janitors had been locked in place by provisions of a union contract but that the contract had been revised in 1965, correcting that situation. The supplemental evidence adduced at the second trial complements the basic facts we found to have been established at the first trial. Upon weighing the supplemental evidence, we find as follows: The supplemental evidence regarding evaluations adds little to the record as it stood at the conclusion of the first trial. Defendant did reprimand its employees who walked off the job to attend a demonstration, but did so across the board, treating white and minority employees the same. Black Electrician B’s spent more time in grade than white Electrician B’s and there were no black Millwright B’s promoted to A before 1969. The individual plaintiffs generally fared no worse in promotions than did white males of comparable tenure and experience. Defendant instituted an affirmative action plan, with particular attention to the Millwright grades, as early as 1968. These findings are consistent with those we made at the earlier trial and directly complement our earlier findings. With respect to the non-statistical circumstantial evidence, we adhere to the basic factual determinations we made at the first trial except to the extent they may be expressly modified by our findings made above on the basis of supplemental evidence. Statistical Evidence By far the greater part of the evidence at this second trial was statistical. Including each of defendant’s sub-exhibits, there were well over one hundred separate statistical exhibits admitted into evidence. Each side presented an expert witness who testified as to the meaning and weight which should be afforded the various statistics. As a preliminary overview, we note that the statistical evidence was aimed at showing promotion trends (and the significance of such trends) in each of the promotion categories, for each of the sex and ethnic groups to which the plaintiffs belonged, and for each of the relevant years. The statistics were variously derived from two slightly different data bases and were analyzed by the respective experts in accordance with three different methodologies. We first summarize the statistics developed under each of the three methodologies. We then make direct comparisons of the various figures, grouping the figures by promotion category. (1) Plaintiffs’ 4Á’s Rule Plaintiffs’ Exhibit 131 both defined and explained the working of the Vs’s rule. According to the rule, as defined by plaintiffs, if the affected class’ representation in a job group is less than Vs’s of the majority representation, then adverse impact is said to exist. Plaintiffs’ Exhibits 124, 125, 126, and 133 apply the Vs’s rule to defendant’s promotion decisions. In each case for which he did calculations, plaintiffs’ expert, Dr. Pearson, first determined a promotion rate for white males and promotion rates for blacks, Hispano-Americans, and females. He then compared the rates and, wherever the promotion rate of the affected group was less than Vs’s that of white males, he concluded that there was an adverse impact for that group. Plaintiffs’ expert performed calculations separately for salaried, hourly-out-of-unit, hourly-in-unit, and hourly to salary promotions. In addition, there was a separate calculation for hourly to salary (officials and managers). In all instances but one (hourly to salary, officials and managers, 1966), he used tape data. He did these calculations for the years 1966-69. The result was a showing of some adverse impact (as defined by plaintiffs’ expert) in some 32 specific instances. It was the expert’s conclusion that this showing established an overall adverse impact. We summarize plaintiffs’ evidence below. For ease of analysis, we have used a tabular form broad enough to encompass like summaries of the exhibits which are to follow. We have replaced the numerical values with three symbols: a indicates a promotion rate for the group which is less than Vs’s the rate for white males, a “ + ” indicates a promotion rate higher than the rate for white males, and “0” indicates a null value (that is, a rate which is less than that of white males but greater than 4/s’s of that rate). Blanks merely indicate the absence of any calculation for a particular field. [See following illustration.] Table (1): Plaintiffs' 4/5's Rule This table is derived from plaintiffs' Exhibits 124, 125, 126, and 133. Table (1) reflects Mr. Pearson’s showing of adverse impact in 32 instances. The table also shows 18 instances where the affected group had higher promotion rates than did white males and 6 instances of null values. There has been no serious challenge to the raw data upon which plaintiffs’ calculations are based, nor has there been any challenge to the arithmetical correctness of the calculations. We find the exhibits in question to be reasonable, credible, and worthy of belief. We, accordingly, find that, in 32 cases out of the 56 cases examined under this method, there is a showing of adverse impact as defined by the %’s rule. (2) Plaintiffs’ Calculations of Statistical Significance Plaintiffs’ Exhibit 132 explains the concepts behind one probability test for determining statistical significance of observed disparities. Plaintiffs’ expert, Dr. Pearson, describes his statistical model as one which measures the “significance of difference between two proportions.” The model begins with the assumption that there is no expected difference in the anticipated performance among the population groups. The hypothesis is then tested by calculating the standard error and the Z statistic. A conclusion about the initial hypothesis may be drawn from the Z statistic. A Z statistic of less than 1.96 (roughly two standard deviations) is within the normal distribution and tends to confirm the initial hypothesis. In the instant context, such a value would indicate that the affected group is receiving an expected promotion rate. A Z statistic equal to or greater than 1.96 occurs less than 5% of the time in a normal distribution and means that the initial hypothesis may be rejected. In the instant context, such a value would indicate that the proportions do not come from the same distribution. Such a statistically significant difference between the groups would be evidence of. ad verse impact. ' Plaintiffs’ Exhibits 127, 128,129, 130, and 134 apply the “statistical significance” test described above to defendant’s promotion decisions. Plaintiffs’ expert performed selected calculations separately for certain salaried, hourly-out-of-unit, hourly-in-unit, and hourly to salary promotions. There was a separate calculation for certain hourly to salary (officials and managers) promotions. Plaintiffs’ expert did calculations based upon selected manual and tape data for certain years between 1966 and 1969. Only statistically significant adverse differences in proportions (Z scores equal to or greater than 1.96) were presented. The result of the calculations for statistical significance was a showing of adverse impact in 14 specific instances. It was the expert’s conclusion that this showing enhanced the demonstration of impact under the 4/6’s rule and that it tended to confirm the existence of an overall adverse impact. We summarize plaintiffs’ evidence below. • For ease of analysis, we have used a tabular form identical to that used previously in table (1). We have again replaced the numerical values with symbols. As the exhibits in question presented only adverse impacts, the only symbol which appears is a ” for each such instance. Blanks merely indicate the absence of any calculation for a particular field. Table (2): Plaintiffs' Test for Statistical Significance Table (2) reflects Dr. Pearson’s showing of adverse impact, as defined by his probability test, in 14 specific instances. There has been no serious challenge to the raw data upon which plaintiffs’ calculations are based, nor has there been any challenge to the arithmetical correctness of the calculations. We find the exhibits in question to be reasonable, credible, and worthy of belief. We, accordingly, find that, in the 14 cases examined under this method, there is a showing of adverse impact as defined by this test. (3) Defendant’s Calculations of Statistical Significance Defendant’s expert witness, Dr. Gollob, explained his methodology at some length, both in oral testimony and through exhibits. His two written reports, Exhibits EEE and BBBB, well explain the concept behind his probability test. He characterized his model as a “lottery system.” The statistical methods employed to establish inferences based upon that model were described as a “normal approximation to the binomial” and as a “hypergeometric distribution.” In brief, the lottery assumed, in the case of salaried, hourly-out-of-unit, and hourly-in-unit promotions, that the individual employees should have an equal likelihood of being promoted and that any employee might receive more than one promotion in each year. Relevant probabilities were calculated by the binomial and are summarized in Exhibit EEE and Exhibits S-l to S-5 which are attached to Exhibit EEE. The lottery assumed, in the case of hourly to salary promotions, that individual employees in the same labor pools should have an equal likelihood of being promoted and that there was a negligible chance that any one employee might receive more than one hourly to salary promotion in a single year. Relevant probabilities for hourly to salary promotions were calculated by the hyper-geometric and are summarized in Exhibit BBBB and Exhibits BBBB-1 to BBBB-9 which are attached to Exhibit BBBB. Under both variants of the lottery system, a factor equivalent to a Z score was derived. In Dr. Gollob’s terminology, the 5% rule of thumb was used to isolate occurrences which are outside a normal distribution. In this context, such values indicate a statistically significant difference between the two groups. Defendant described those cases where the Z score equivalent was less than 1.96 as being inside the bounds of chance and described values of 1.96 or higher as being outside chance. In addition to the summary defense exhibits already noted, Defendant’s Exhibits EEE-1B to TTT-3 present 108 separate applications of the “statistical significance” test described above. In applying the lottery test to defendant’s promotion decisions, defendant’s expert performed separate calculations for all salaried, hourly-out-of-unit, hourly-in-unit, and hourly to salary promotions. Dr. Gollob used all the manual data and all the tape data available for every year from 1966 to 1971. He performed several separate calculations for hourly to salary promotions, one ignoring the relevant labor pools, others taking the labor pools into account, and additional ones which focused on hourly to salary (officials and managers) by labor pool. He presented his results in every case, showing not only instances of statistically significant adverse promotion rates, but also showing instances where there were null values (within chance) and positive values (statistically significantly higher promotions for the minority groups than for white males). The result of these calculations for statistical significance was a showing of adverse impact in 20 specific instances. The calculations also show 23 instances where the minority groups received a statistically significantly greater promotion rate than did white males and 197 instances of null values (variations of promotion rates within chance). Defendant’s expert concluded that this showing strongly supports the belief that neither blacks nor Hispano-Americans were adversely affected by Martin Marietta’s promotion policies. Dr. Gollob concluded that the overall pattern of results also supports the belief that females were not adversely affected. We summarize defendant’s evidence below. For ease of analysis, we have used a tabular form identical to that used previously in tables (1) and (2). We have replaced numerical values with symbols: a ” indicates a promotion rate for the group significantly lower than the rate for white males, a “ + ” indicates a promotion rate significantly higher than the rate for white males, and a “0” indicates a null value, that is, a rate which lies within the bounds of chance. Blanks merely indicate the absence of any calculation for a particular field. Table (3) reflects Dr. Gollob’s showing of adverse impact, as defined by his probability test, in 20 specific instances. The table also shows 23 instances where the affected group had significantly higher promotion rates than did white males and 197 instances of null values. Table (3): Defendant's Test for Statistical Significance This table is derived from defendant's Exhibits S-l, S-5, BBBB-1, BBBB-8, and BBBB-9 There has been no serious challenge to the raw data upon which defendant’s calculations are based, nor has there been any challenge to the arithmetical correctness of the calculations. We find the exhibits in question to be reasonable, credible, and worthy of belief. We, accordingly, find that, in 20 cases out of 240 cases examined under this method, there is a showing of adverse impact as defined by this probability test. We also find that, in the great majority of instances, there is no adverse impact. (4) Direct Comparisons of Plaintiffs’ and Defendant’s Figures Having summarized the mass of statistical data in a standardized tabular form, it becomes a simple matter to make direct comparisons among the several exhibits. Appendix II, attached to this opinion, is a set of four charts which make those cornparisons for salaried, hourly-out-of-unit, hourly-in-unit, and hourly to salary promotions respectively. We shall refer to those Appendix II charts in more particularity later on in the course of this opinion. For now we simply note that there are certain differences in results depending upon whether tape or manual data were used and whether (for hourly to salary promotions) labor pools were taken into account or not. In addition to those differences, which turn upon which raw data were selected, there are further differences which result from the use of one or another of the three methodologies chosen. Surprisingly, the major dichotomy is not between the plaintiffs’ two methodologies on the one hand and the defendant’s methodology on the other. Rather, the chief distinction is between results obtained by plaintiffs’ 4/5’s Rule and those results obtained by either of the tests for statistical significance (plaintiffs’ or defendant’s). We have already made certain observations concerning the data bases (see note 13, supra) and labor pools (see pages 593-594, 595). We have likewise noted the salient features of each of the three methodologies used by the expert witnesses (see note 14, supra; pages 598-606) and made the factual finding that the various calculations under each method are internally sound. The determination as to which of the data bases, labor pools, or statistical methodologies are the more probative is a legal conclusion which we shall address presently. Ill The well known case, McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973), supplies an example of the kind of showing a plaintiff must make to establish a prima facie case. 411 U.S. at 802, 93 S.Ct. 1817. The importance of McDonnell Douglas lies not in its specification of certain discrete elements of proof “but in its recognition of the general principle that any Title VII plaintiff must carry the initial burden of offering evidence adequate to create an inference that an employment decision was based on a discriminatory criterion illegal under the Act.” Int’l Brotherhood of Teamsters v. United States, 431 U.S. 324, 358, 97 S.Ct. 1843, 1866, 52 L.Ed.2d. 396 (1977). That is to say that the plaintiff must show the occurrence of actions “from which one can infer, if such actions remain unexplained, that it is more likely than not that such actions were ‘based on a discriminatory criterion illegal under the Act.’ ” Furnco Const. Corp. v. Waters, 438 U.S. 567, 576, 98 S.Ct. 2943, 2949, 57 L.Ed.2d 957 (1978) (quoting from Teamsters, supra). There is a wide range of evidence which, in differing factual situations, might serve to satisfy that burden and establish a prima facie case. See, McDonnell Douglas, supra, 411 U.S. at 802 n.13, 93 S.Ct. 1817. The particular method suggested in McDonnell Douglas was never intended to be regarded as rigid or ritualistic. The central focus of the inquiry is always whether the employer is treating some people less favorably than others because of their race, color, religion, sex, or national origin. Furnco, supra, 438 U.S. at 576, 98 S.Ct. at 2949. The mandate of the Tenth Circuit contains additional guidance for further proceedings in the instant case which we are, of course, bound to follow. After noting that the elements of a prima facie case are flexible and varied, the appellate court did give an intimation of at least one set of elements which would serve to make out a prima facie case in the instant context. Where a plaintiff has shown “[1] that he is qualified, he need only show [2] a discriminatory impact, and [3] that he was among the class of employees who could have been considered for promotion.” Rich, supra, 522 F.2d at 348 (numerals added). IV We conclude that five of the seven pláintiffs have failed to establish a prima facie case of employment discrimination under Title VII. Of the seven plaintiffs, only Ms. Rich and Mr. Chappel have met their initial burden of proof. Plaintiffs have patterned their case in the fashion suggested by the Tenth Circuit. They have urged that they each were qualified for promotion, that there were available positions into which they could have been promoted, and that these qualified plaintiffs were denied available promotions because of the discriminatory intent or impact of defendant’s promotional policies. We make our conclusions of law on these three elements of plaintiffs’ claims, in turn, and in the same order in which we have made our findings of fact. A Lengthy tenure together with knowledge and experience, in the context of this case, will establish qualifications. Rich, supra, 522 F.2d at 348. In this second trial, the parties have argued whether the appellate court went on to hold that each of these plaintiffs was, in fact, qualified for promotion or whether it left the factual question open. We conclude that the court did, in practical effect, so hold. Once qualification is understood primarily in terms of longevity, it must follow, on the facts of this case, that these plaintiffs are qualified for promotion. B It would appear at least as firmly established that these plaintiffs were among the class of employees who could have been considered for promotions. The Tenth Circuit has held that there is no necessity to point to any specific opening within 90 (or 210) days prior to plaintiffs’ having filed charges with the EEOC. Plaintiffs have alleged a continuing violation, hence the timing is inconsequential. Rich, supra, 522 F.2d at 347, 348. It is apparent then that two of the three elements necessary to at least one relevant form of prima facie case are present here. The only item left for plaintiffs to show is “discriminatory impact.” Given the presence of the other two elements, if the plaintiffs, or any of them, can show a “discriminatory impact,” then a prima facie case will have been established. We now turn to this crucial element. C In the instant case, “discriminatory impact” could be established by a showing of discriminatory intent or by a showing of significantly discriminatory results of facially neutral practices. It may fairly be said that plaintiffs here have sought to show both and have thus attempted to make out a case both under the theory of disparate treatment and the theory of disparate impact. It is not entirely clear that the Tenth Circuit, in 1972, used the word “impact” in the specialized sense it has subsequently gained. Defendant goes so far as to urge that the court of appeals has since questioned its own holding in Rich for that reason, citing, Olson v. Philco-Ford, 531 F.2d 474, 478 (10th Cir. 1976) for that proposition. We conclude that defendant misconstrues Rich. When the Tenth Circuit used the term “impact,” it is now clear that it must have meant that it was incumbent upon plaintiffs in Rich to show such an impact either by showing motive (to establish disparate treatment) or by showing significantly adverse results of neutral practices (to establish disparate impact). We conclude that such an understanding of “impact” is certainly consistent with the mandate in this case and is sufficient to dispel any asserted conflict with subsequent cases. Indeed, such an understanding would seem the only logical choice. It explains not only the comments in Olson, supra, but also the directions in Rich itself that, on remand, this court reconsider the evidence and “determine whether the plaintiffs have proven a prima facie case.” Rich, supra, 522 F.2d at 347 (emphasis added). It is quite clear that the court of appeals never intended a prima facie case to be established whenever a plaintiff merely shows that he or she was qualified for a promotion that was, in fact, given to a qualified member of another sex or ethnic group. Defendant’s concerns on this score are, thus, unfounded. As noted, we earlier (in the first trial) found the circumstantial evidence on this key element of “discriminatory impact” to be unconvincing. Though the court of appeals has directed this court to reconsider this evidence, there is no hint that it determined any of our underlying findings of fact to have been clearly erroneous. Building upon the factual findings made at the first trial, we shall reevaluate the circumstantial evidence as to each plaintiff. We likewise earlier found the statistical evidence on this crucial element of impact to be unpersuasive. There was error here, arising from the fatally narrow scope of discovery which, in turn, caused the statistics offered at the first trial to be unsatisfactory. This error frustrated the search for truth and left doubts as to the merits of plaintiffs’ claims. Rich, supra, 522 F.2d at 343. The Tenth Circuit has directed this court to consider additional statistical evidence and then determine anew whether plaintiffs have established a prima facie case. Indeed, it is apparent that the statistical evidence is of utmost importance in this case and that [particularly in light of the contention of the defendant and the findings of the [trial] court that the circumstantial evidence was ambiguous, it became the more necessary for plant-wide statistics and facts to be obtained and presented, for they very likely would prove crucial to the establishing or failure to establish a prima facie case. Rich, supra, 522 F.2d at 344 (footnote omitted) (emphasis added). Disregarding the defective statistics of the first trial entirely, we shall consider the relevant additional statistical evidence as it relates to each plaintiff. It is necessary, however, preliminary to considering the statistics, to determine which of the data bases, labor pools, or statistical methodologies are the more probative. There was some difference in results depending upon whether manual or tape data were selected. Both sets of data were supplied by defendant. Defendant concedes that the tape data does not include information on certain employees who had worked at Martin Marietta during the relevant period (1965-71) but who may have left the company before it changed over to computer record keeping (circa 1971 — 72). Defendant explained its reasons for so having set up its tape data in the ordinary course of its business. Defendant also explained the difficulties which stood in the way of reconstructing the tape data specially for this lawsuit. Plaintiffs do not challenge defendant’s good faith on this score, nor do we. Indeed, the variance in results produced by choice of data base is not predictable nor is it one-sided. Sometimes the manual data will show an instance of adverse impact where tape data shows none, sometimes it is the tape data which shows an instance of adverse impact while the manual does not. Despite its justifiable explanation for the variance in data bases, we nonetheless conclude that defendant must be bound by the data it produced and that defendant must bear the risk of uncertainty created by its different data bases. We will resolve any possible doubts in plaintiffs’ favor. We have considered any data source, manual or tape, that reflects adverse impact, even though the other data source might not show a corresponding instance of adverse impact. There was a rather substantial difference in results depending upon whether (for hourly to salary promotions) relevant labor pools were used. Statistics, such as those produced by both parties at the first trial and by plaintiffs at the second trial, which ignore the labor pools present an entirely different picture than do the statistics which take the labor pools into account. We have found, as a factual matter, that labor pools do exist at Martin Marietta. The evidence on this score is abundantly clear, overwhelming, and uncontradicted. It must follow, both legally and logically, that the proper comparisons must be between the racial, sex, and ethnic composition of results under defendant’s promotional policies and the racial, sex, and ethnic composition of its qualified employee population in the relevant labor pools. See, Hazelwood School District v. United States, 433 U.S. 299, 308, 97 S.Ct. 2736, 53 L.Ed.2d 768 (1977). Where, as here, the salaried workers are, for the most part, professionals performing technical jobs; where, as here, the hourly workers include a predominance of office and clerical, skilled and unskilled workers in the lower grades and a predominance of technical workers in the higher grades; and where, as here, it is clearly proven that the higher grade technicians are the more likely to be promoted from hourly to salary, it cannot realistically be maintained that the qualified employee pool for these promotions consists of the undifferentiated mass of hourly employees. On the clearly established facts of this case, it simply cannot seriously be argued, for example, that a janitor at Martin Marietta and a Millwright A (though both hourly workers) have equivalent qualifications for promotion from the hourly ranks to the professional positions held by defendant’s salaried employees. We conclude that the more probative statistics are those which take labor pools into account and we are compelled to reject those statistics which ignore the labor pools. Finally, there were differences in results depending upon which of the three methodologies was used to define an instance of adverse impact. In this general regard, it has been noted that the crux of the current method of proof in discrimination cases is the presentation of percentage differences sufficiently substantial to suggest bias. Note, Beyond the Prima Facie Case in Employment Discrimination Law: Statistical Proof and Rebuttal, 89 Harv.L.Rev. 387, 393 (1975). The problem comes in deciding, given an observed difference in percentages, just what degree of disparity constitutes a significant or substantial disparity. This is a question which has not clearly been resolved. Id. at 393 n.26; see also, Note, Employment Discrimination: Statistics and Preferences Under Title VII, 59 Va.L.Rev. 463, 477-80 (1973). Indeed, it can be one of the more perplexing questions in a case such as this, tending to leave the parties as well as the court in the unsatisfactory position where unguided intuition has, seemingly, been the touchstone. Recent cases have taken the lead in using statistical techniques of one sort or another and, in so doing, have provided sorely needed clarification in how to approach and resolve this problem. See, Hazelwood School District v. United States, 433 U.S. 299, 308-09 n.14, 311 n.17, 97 S.Ct. 2736 (1977) (using the binomial distribution in a Title VII case); Castaneda v. Partida, 430 U.S. 482, 496-97 n.17, 97 S.Ct. 1272, 51 L.Ed.2d 498 (1977) (using the binomial distribution in a jury selection case); Otero v. Mesa County Valley School District, 568 F.2d 1312, and footnote at 1316 (10th Cir. 1977) (recommending consideration of the binomial methodology of the Hazelwood case in an analogous situation). In the presentation of evidence in the instant case, both parties have anticipated these recent trends in the direction of somewhat more sophisticated statistical techniques. See, e. g., the authorities collected, supra; see also, Shoben, Differential Pass-Fail Rates in Employment Testing: Statistical Proof Under Title VII, 91 Harv.L.Rev. 793 (1978). Both plaintiffs and defendant have presented the court with calculations performed under one or more methodologies tailored to providing a principled basis for interpreting the significance of an observed difference in raw percentages. Defendant relied squarely upon the more important of the recent cases, using a hyper-geometric distribution and also the binomial distribution which has been expressly approved in the Hazelwood, supra, Castaneda, supra, and Otero, supra, eases. We shall compare each of plaintiffs’ two methodologies to the method selected by defendant. Plaintiff relied, first, upon the %’s Rule. It should be noted at the outset that this rule is not, strictly speaking, a “statistical” method at all. It is no more or less than a straightforward comparison of raw percentages, with a mechanical rule of thumb then applied in a regular manner so as to substitute uniformity for the “flexibility” (see, note 27, supra) which has characterized other such ball park comparisons. It has the virtue of simplicity and, as it has been uniformly adopted by four federal agencies, it has the companion virtue of providing the potential for predictability of results in discrimination cases. The rule should not, however, be given deference beyond its due. Its narrow primary use is clearly set forth in the explanatory information which emphasizes that the rule does “not address the underlying question of law. [It] discussfes] only the exercise of prosecutorial discretion by the Government agencies themselves.” 43 Fed.Reg. at 38291. (1978) (footnote omitted). Again, it is noted that “rule of thumb” is “not a legal definition of discrimination, rather it is a practical device to keep the attention of enforcement agencies on serious discrepancies. . . .” Id. (emphasis added). It is apparent that, while by no means conclusive, the 4/s’s Rule might well produce some evidence of discrimination. In determining how much weight to give such evidence, we are guided, in part, by the observation that “statistics are not irrefutable; they come in infinite variety and, like any other kind of evidence, they may be rebutted. In short, their usefulness depends on all of the surrounding facts and circumstances.” Teamsters, supra, 431 U.S. at 340, 97 S.Ct. at 1856, 1857 (citation omitted). In the context of the case at hand, we could well say that, if the Vs’s Rule calculations were all the evidence we had before us, we might well find them to be highly persuasive and we might seriously consider using the benchmarks developed by the government agencies to help interpret the figures. But, while it may supply an inference of discrimination, the inference of the Vs’s Rule (resting, as it does, upon untested intuition) is not a strong one. And, while the Vs’s Rule may produce some evidence of discrimination, we are confident that the binomial distribution (approved by the Supreme Court, and the Tenth Circuit, and used by defendant) produces much better evidence. On the particular facts and circumstances of this case (not the least of which was the ability of defendant’s expert witness